Impact of Dropsonde Observations on the Predictability of Landfalling Atmospheric Rivers Minghua Zheng, Bruce Cornuelle, Luca Delle Monache, F. Martin Ralph Center for Western Weather and Water Extremes Scripps Institution of Oceanography, University of California San Diego, USA Center for Western Weather And Water Extremes

Introduction Results for IOP7: Precipitation Forecast Differences Landfalling Atmospheric Rivers (ARs) provide between 30-50% of precipitation, but can Stage IV Precip. From February 03-04 NoDROP Precip. From February 03-04 (F12-F36) also cause major flooding events in the western U.S. Accurate forecasts can improve water management decisions. Sparse observations over the Pacific have limited the improvement of forecast skills for the western U.S. due to the poor upstream initial conditions. While numerical weather predictions reap the benefits of satellite data over the oceans, those data poorly represent the low-level circulation and the vertical structure of water vapor in ARs. In the winters of 2016, 2018, and 2019, fifteen aircraft reconnaissance missions were carried out targeting AR conditions over the eastern Pacific. More than 900 dropsondes measured vertical profiles of wind, water vapor, temperature, and pressure. In this work, paired data assimilation experiments with and without these dropsondes are conducted using the Weather Research and Forecasting (WRF) model and the Community Gridpoint Statistical Interpolation (GSI) data assimilation system. The impact on the forecast skills of precipitation over the western U.S. is evaluated by comparing the simulations from NoDROP Precip. Error Impact: WithDROP minus NoDROP paired runs for all fifteen Intensive Observation Periods (IOPs). IOPs Dates IOPs Dates

IOP1 02/14/2016 IOP9 02/28/2018 0 10 20 30 40 50 IOP2 02/16/2016 IOP10 02/02/2019 GPM Precip. Rate (mm) NEXRAD Reflectivity IOP3 02/22/2016 IOP11 02/11/2019 IOP4 01/27/2018 IOP12 02/13/2019 A list of IOPs and the corresponding dates. IOP5 01/29/2018 IOP13 02/24/2019 IOP6 02/01/2018 IOP14 02/26/2019 IOP7 02/03/2018 IOP15 03/01/2019 IOP8 02/26/2018 Sounding Observations WRF Configuration 50oN • Resolution Results for All IOPs: Impact on Precipitation Skills ― 9 (d01) / 3 km (d02), 48 levels, All Lead Times (day 1 to day 5) 10 hPa top 40oN • Initial and boundary conditions ― 6 hourly GFS 0.25 x 0.25 º forecasts 30oN • Physics ― One-way nested 20oN ― New Thompson scheme → Improvement ― Yosei University PBL scheme ― Grell 3D cumulus scheme 10oN ― RRTM long wave radiation ― GSFC short wave radiation scheme o o ― Two-hour digital filter 150 W 120 W

Data Assimilation Experiments Set-up ← Degradation

• GSI Hybrid 4DEnVar RMSE reduction: RMSE of NoDROP minus RMSE of WithDROP. Positive: improved skill in WithDROP ― Static background error covariance Error-Diff Correlation: Correlation of NoDROP errors with NoDROP-WithDROP differences. parameter: 0.25 Positive: improved spatial correlation in WithDROP. ― Ensemble: downscaled 20-mem GEFS All IOPs + 20-mem CMCE ― Background: 9-hour forecast from WRF • Assimilated Observations ― Conventional data in NCEP PrepBUFR ― Satellite derived wind in BUFR format Improvement → Improvement ― GPS RO refractivity in BUFR format AR Recon Dropsonde • Paired experiments Optimized WRF Sounding ― WithDROP: dropsondes assimilated ― NoDROP: dropsondes not assimilated • Verification domain and variables o o ← Degradation ― Common domain: 20-50 N, 110-125 IOP IOP W Note: IOPs are defined as consecutive cases if two adjacent IOPs are within 4 days. The Arrows in the ― Accumulated 24-h precipitation using above plots point from the first to the last IOP in a group of consecutive cases. Stage IV 4-km data as the ground truth (mm) Stage IV 24-h precipitation over verification Conclusions and Future Work domain at 12Z, 3 February 2019. Impact of dropsondes on precipitation skills over western U.S. • The median RMSE reduction in WithDROP runs are positive for most of the lead times except for 72-h and Results for IOP7: IVT Differences 120-h. Precipitation RMSE skills are significantly improved at 24-h and 96-h lead times. The spatial correlation are significantly improved at all lead times. Initialized at 00Z, 3 February 2018 • Eleven out of fifteen IOPs show that the median value of RMSE at all lead times are improved. Fourteen out of fifteen IOPs show the spatial correlation at all lead times are significantly improved. • Larger improvements are consistently found after the first IOP of a group of consecutive cases, i.e., IOP2, IOP5-7, IOP9, IOP12, and IOPs 14-15.

Future work • Evaluating the data impact with satellite radiance assimilated. • Investigating the physical processes that might be associated with positive or negative impacts. • Choosing different verification domains for IOPs not making landfall at western U.S., such as IOP3 and IOP8.

(a) Forecast error (filled, kg m-1 s-1) of IVT in NoDROP run at 24-h forecast lead time. Acknowledgements (b) Same as (a) but for WithDROP run. Black contours in (a) and (b) are analyzed IVT from The AR Recon were a joint effort of UCSD/SIO/CW3E, NOAA/NWS, ECMWF, NRL, and, the Air Force. Part of ERA5 reanalysis data valid at 00Z, 4 February 2018. this work was finished during Dr. Minghua Zheng’s visit at NOAA/NCEP funded by UCAR’s CPAESS program. Contacts: Luca Delle Monache ([email protected]) and Minghua Zheng ([email protected])